Press "Enter" to skip to content

Ces machines qui décodent nos expressions faciales | Mohamed Daoudi | TEDxIMTLilleDouai


good evening everyone i am very very happy
to be with you tonight to talk about
things that concern us all about your
face of your emotions which preparing him
this presentation there was to an article that
just released in forbes magazine
who quotes the big ones I ‘ll go trendy
at the technological level that go somewhere
share changing the world and among these
big trend
we are talking about the rise of
artificial intelligence
so this ability to endow
intelligence machines and this for
two main reasons the first one
has increasingly makes machines
which are becoming faster
and then the second reason is we have to
more and more to give big bases
of data that allows us to do this
we did apprenticeship
so this computing power and has
given today then allow to
our computers doing things
interesting enough they can put it
to computers to see for example
today you have systems of
face recognition that works
very very well we use what we
call learning dating you have
computers that are capable of you
listen to you speak and see gauge your
emotions will understand
so certainly on the terms in
I’m going to talk to you today
what today’s computers
are able or to be able to
understand your emotions
so why the emotions interests
everyone because the emotions
are part of our everyday life
the emotions so go not had not put
finally social human interaction
is that you very important the
emotions have a very important role on
our health and well-being
and then sometimes they even guide
us our purchases I would almost make
attention because some even
would like to use them to incite you
to buy
so pay attention to that and finally
in our predictions emotions play
a very very important role for all
these reasons there actually the emotions are
something important to understand
but next to that what we have
today we are more and more
connected we have machines apart
smartphones computers
connected to these machines
but these machines are becoming more and more
very very intelligent and very powerful
but do not have that ability there of
to interact with us a little bit
more pull human is a way of their
have said badly if he is to inject a little
more emotion so the goal is
just maybe tomorrow you’ll have
maybe a car and I think I
am sure and certain in 2 3 years you
are going to have cars equipped with
small cameras that will analyze your
face from this analysis there they
are going to understand maybe you are in
a stressed state and maybe triggered
a little softer music and see
offer you services
it’s going to be in our homes is from
more and more connected maybe tomorrow
you are going to have in your
fridge that’s going to be getting
more connected a small camera that
had analyzed your emotional state and
you might suggest ddd menus in
function of your state at the moment t
and that’s what that’s about
getting ready but the study of
emotions do not date from today
it already dates from charles darwin who
tried to understand a little how to
through our face and our body we
happens somewhere sent some
emotions to others he became interested in
that not only for us finally
human beings but also for
animals
so that dates from very, very far
so what I ‘m going to tell is
past it’s true that when I was invited
here at tedx i’m still in fact
interested in talking about what I’m doing at
research level
and when I was told it was douai so
I said potato to Lille Lille
douai 1 saturday night
finally while I take my car etc
but let’s think and working on
this presentation they did the really a
very good choice to organize it was dixie
ca do and because you will see
the study of the analyzes explodes the
facial expressions was born almost here
at 12 o’clock the head of the great oak
French neurologist who was born in Bolts
on bert and who studied
school here to adu and who went to
taken after to paris to celebrate the studies
of medicine after he returned home
him in boulogne-sur-mer and oak this
that he has what he tried to do when
you see on your left
he tried to understand we are stimulating a
number of muscles alone face
what is happening in terms
expression and realized
observation is that stimulated
muscles in the wrong is accurate on
face they arrived generates to center
of facial expressions
for example the joy so he was doing that
I gather two passions so I you
do not recommend doing this at home
with the readers it and what is it
it’s dangerous
so that’s what you’ll see these
work of the leader anne impact
impressive about what we do
today then later more carved
other work and it’s really
quite important work that has a
impact on the
what we are doing today a new
analysis of the heroes on the expression that’s
paul ekman paul ekman who who also
worked on the work of the oak is
that’s interesting but paul is that
badly he was interested in understanding
certain universality our expressions
facial in what he did so that’s
an American civilian group what the
makes the parties to papua new
Guinea
an expedition is exhibited in a tribe
who had never actually seen culture
manage foreign aliens and there
look at these people how they
actually arrive from a photo of
people who were rather in the west
and what they come to understand their
facial expressions the cheek enough
and if we finally count that these
people arrived to understand well in
do all our ways our our joy
our sadness and so is what he has
does he established from douala and
has established a theory saying that he
ya star or facial expressions that
are universal who are we going
find later at our level
analysis of two faces who are we going
find joy sadness anger
disgust and fear so we have six who
are fundamental and the rest whatever
share its combination of these six
bath expression and so from
that time the work of researchers
as in my case on the plan
computer vision was to see
that it was possible to offer 2.2
from algorithms to programs
computer capable from a
video where an image to say this
expression this face there it corresponds to
this or that facial expression
so that’s all work and the
difficulty is that when a camera at
look at your face she sees
numbers and see not anything else we
see that numbers
so all the work behind these two
what can we understand from
of those numbers with difficulty
is that if you ask me from you
write a port one for a person who
is in front of a camera to analyze his
face I can do it I can do it
because you simply have to analyze the
the morphology of the face can be done
some pretty interesting things for a
only person but I ask you
to describe
and propose a computer program
for millions of people the
lucky things because we have a
different morphology we sometimes have
slightly different different expressions
and herat and so you have to think
otherwise we can not write a
computer program like that to think
otherwise what is it
come back my starting point my point of
it was that today we have a
computing power we have data the
give is so that’s what we will do
here is what we will do here with
all the difficulties we are going
to meet is that a person like
here you see it’s princess k is
here he is in different positions
different grimaces and we even have
If extra right here with great
wines are even more exaggerating so he
must be considered in fact what we
calls all these variability of
data so what are we gonna do
build what we call the
benchmark actually benchmarks and
big databases and we’re going
collect a lot of data a lot of
data that fits all these
facial expressions and we will try
learn
we will learn from that
exactly what did when
teach a child we will introduce him
a face
we are going to tell him that corresponds to anger
this face there is the surprise and
it is so we will try to learn his
different facial expressions and for
we need to give it we see
today gives it and we can
control we can do things
ourselves rebuild this tube not get
give or just sometimes
we go directly to youtube
we have billions of data to our
disposition with this type of expression
so we will learn we will learn in
using learning techniques
I would like an example here to go back to
what you find in any system of
recognition of facial expressions
first of all the data one can have
a little treatment to treat these
data there but the kernel part the
more important to the most part
important fundamental this part there
so a machine that will learn we will
present them with data we will tell him
it’s a smile, it’s like
anger
and then we’re going to turn these
machines there for several days
even two days when finished we have
a system that is capable of
recognize our expressions
facial expression vassal facial
with all the different variability
which I told you about earlier
so I would like an example in fact of
to confess to do today
so here I have a little video to you
so show there is an example of this
we know how to do today we have here a
system that we developed we
same who who actually learned all
these facial expressions and the person
who is in front of the camera now
it’s really a real time system
what you see is the actual rate
that is to say that you are before
camera and that it works so the person
he will do his language of expression and the
system he’ll recognize that she’s there
the expression that corresponds to this
no one there you see here at the level of
chama here on the face the system and
they detect other types
of facial expressions but is still
one that is dominant at the level of
face
so that’s it’s less students
zapped a man space who developed this
system there it actually pretends
they never angry it’s always
he is someone who is always happy
like what you see here so the way
there and then what will try to
show it too and that’s what it is
showing that the system
is very robust
when the person moves his head and
that ‘s it’s very important in this
type of system to have this type of sound
call of robustness to the movement of the
head
and then there they exaggerate a little it
play the actor the ai really he does it
good
so then the level of the
demonstration what we show you too
this is also important
because tomorrow it’s a subtle system
it can be installed on a smartphone
or a tablet so he will be able to
light conditions that changes so we
needs systems capable of holding
actually account this different changes
so there the system
he adapts he is able to see you
there even took off his glasses and the
system he is always able to
recognize the expression of this
no one
so that’s the ball so that’s a
then this type of system
you will soon have it in fact on your
smartphones or tablets it’s
things in fact today whose area
as soon as possible maybe even bicycles
cars that interests us too
and indirectly many of these
work there actually shame on the impact
bullish in the medical field and
make sure that we really care about that
we have a lot of work in the field
of medicine or an example that’s what
what is called pain today
it is something very important
you see pain analysis by
unfortunately for some people
the only way to express their pain
it’s through their faces and so
doctors need a salary
even cameras to analyze
the expression of the person is detected
the moment of pain or indeed
able to measure the intensity of the
pain
and then finally I mean actually
work who me who really
touched and then I spent a lot of
time at ch here at the southern island with
the experts in facial surgery here
we had people following for example
of a stroke and especially with paralysis
facial and ask them to simply
to make a little smile it becomes
really something hard and so
the doctors at that level the goal
it’s 2
his call to his mistresses and face
because it’s important that the
nobody can find her in fact her
almost original smile so they go
to inject or to inject
call to surgery and at this level
there our role as a researcher in
IT is to put in the
disposition of his doctors there of
tools that will allow them to evaluate
to measure what is called symmetry
of the face before the operation and after
the operation so we asked the
patients to come during before
the operation and we canet them then we
was still very far away because
we have this case themselves their faces
in 3d so their smile you were in 3d
and so we measured what’s called the
Semitour their face before the operation
and then we redid the same
thing they were asked to come to ch
finally here at the service to analyze their
face and we have another arrangement of
tools now that allow for
measure this type of its symmetry call
of the face before and after the operation
here are some some ideas actually this
than today what can
do with
with our faces and then I will finish
by saying the following thing while preparing
this talk about emotions etc
I realize that in fact there is
feeling maybe we were in the process of
to robotize in fact the human being with
all that is what is the ski and
mechanization of tasks and then
now we realize that we have to
actually humanize the machines here
thank you all
[Applause]
Please follow and like us: